A Statistical Perspective on Retrieval-Based Models

Authors: Soumya Basu, Ankit Singh Rawat, Manzil Zaheer

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5. Experiments
Researcher Affiliation Industry 1Google, Mountain View, USA 2Google Research, New York, USA 3Google Deep Mind, New York, USA.
Pseudocode No The paper describes its methods through prose and mathematical formulations, but it does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code No The paper does not provide any explicit statements about releasing source code, nor does it include links to code repositories for the described methodology.
Open Datasets Yes CIFAR-10", "Image Net dataset", "ALIGN (Jia et al., 2021)
Dataset Splits Yes We randomly generate a train set of size n = 10000 in a 10-dimensional space... and perform a 10-fold cross-validation.", "We randomly partition the data into a train set of size n = 10000 points and remaining 2000 points for test. We do a 10-fold cross-validation.", "We use the standard train-test split with n = 1281167 training and 50000 test examples.
Hardware Specification No The paper mentions computational cost and model sizes but does not provide specific hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies No The paper mentions models and optimizers (e.g., 'Adam optimizer', 'Mobile Net V3') but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes For solving the local ERM, we fine-tune a Mobile Net V3 model, which has been pretrained on Image Net, on the retrieved set using Adam optimizer with a linear decay schedule.